I always assumed analogies were explicitly designed as a mechanism for testing vocabulary, not reasoning. I don't understand how analogies would have any value in testing reasoning skills. I can't imagine an analogy question that isn't either obvious or subjective.
The whole point of analogy problems is to test reasoning skills. Specifically, logical skills, inferences, categorization, and so forth. The vocabulary test nested within the analogy test is incidental. It creates a bivariate challenge (vocabulary + categorization), which is not necessarily an invalid test. It's just not a pure test of analogies, and it's also duplicative of the vocabulary portion of the test.
In re the value of analogies, there are some computer scientists and philosophers who believe analogy-drawing is the irreducible core of higher cognition. I'm not sure I'd go that far, but Douglas Hofstadter has come close:
I've actually experienced this years and years ago on my grandmother's machine. It played Fur Elise because it had a short. We needed to figure out the name of the song in order for tech support to help us identify what it meant.
